JOAQUIM BENTO DE SOUZA FERREIRA FILHO* External Adjustment, Production Subsidies and Agricultural Growth in Brazil INTRODUCTION

Transcription

1 JOAQUIM BENTO DE SOUZA FERREIRA FILHO* External Adjustment, Production Subsidies and Agricultural Growth in Brazil INTRODUCTION At the beginning of the 1980s, the Brazilian economy experienced a drastic change in its pattern of development. The appearance of the external debt crisis, together with the continuing effects of the world oil crisis of the late 1970s, saw the end of the import substitution model, which had been financed by massive external capital inflows. The interruption of these financial flows, together with increased international interest rates, made it necessary for Brazil to rely increasingly on domestic saving to service the external debt and to meet growth targets. For the farm sector, this saw the end' of agricultural subsidies and the introduction of indexation on rural credit contracts. The drastic reduction in agricultural support, in a very short time period, lowered expectations about the growth possibilities of the agricultural sector. That fear was sustained by the fact that agricultural subsidies amounted, in 1980, to about 22 per cent of agricultural GDP, or about 2.4 per cent of total GDP. Corresponding figures were, respectively, 14.0 and 1.41 per cent in 1981, 16.0 and 1.4 per cent in 1982, 10.0 and 1.1 per cent in 1983, going down to 0.78 and 0.09 per cent in 1985 (Shirota, 1988). The effects of the reductions were even more drastic than they might appear since total GDP experienced a fall of 8 per cent between 1980 and But the Brazilian agricultural sector, unexpectedly, grew in that period, despite the fall in agricultural subsidies. The real GDP contributed by agriculture expanded by about 10 per cent between 1981 and 1984, and by about 20 per cent between 1980 and In fact, Brazilian agriculture not only grew throughout the entire decade of the 1980s, but its rate of growth was faster than that of the industrial sector (Ferreira Filho, 1996a). This phenomenon is exactly contrary to what would be expected by looking at the issue in a partial equilibrium framework. In that case, the reduction of agricultural subsidies would cause a movement of the supply curve to the left, thus reducing agricultural output. It will be argued here that the problem is too complex to be analysed in partial equilibrium models. The amount of subsidies involved is large enough to have generated macroeconomic effects which could have been dampened the microeconomic ones. The purpose of this paper is to evaluate the problem in a general equilibrium framework, using a computable general *Escola Superior de Agricultura Luiz de Queiroz, Universidade de Siio Paulo, Brazil. 401

2 402 Joaquim Bento de Souza Ferreira Filho equilibrium model (CGE) to perform counterfactual simulations in order to analyse the linkages and feedbacks relating to the fall in agricultural subsidies and the simultaneous agricultural growth. THE CGE MODEL1 The CGE model used in this study is based on the structure of the RUNS model (Burniaux et al., 1990). However, a number of important modifications were made to the model to make it match the Brazilian economy in the 1980s and to focus on the problem at hand. To avoid confusion, the model presented here is the 'Megabnis' version. In the following summary, the model presented is a static one, with the simulation for each year being linked to the others through exogenous growth rates attributed to factors and capital stocks. The economy is divided into two distinct sectors: rural and urban. The rural sector has 11 activities, producing soybeans, sugarcane, corn, coffee, rice, cotton, wheat, other agricultural products, livestock, milk and poultry. Urban sector efforts are separated into seven typical activities (transport, engineering, fertilizers, chemicals, energy, services and others), but also covers 10 agroindustries dealing with the processing of coffee, sugar (including alcohol), rice, wheat, fibres, vegetable oils, meat (excluding poultry), poultry and milk, plus the feed-producing industries. There are, thus, 28 productive activities in the model, each dealing with only one 'product'. There are four institutions in the economy - rural and urban families, enterprises (investment) and government - and three primary factors of production: labour, capital and land. Only the agricultural sector utilizes land. Owing to their great degree of homogeneity the domestically produced agricultural products are assumed to be perfect substitutes for imports. Note that this does not refer to processed goods. The 'small country hypothesis' is also used so that agricultural tradable prices are defined by the exogenous world prices and import or export tariffs, while agricultural non-tractable prices are defined by excess demand in each market. For the period considered in the analysis (which is the first half of the 1980s) there are only three agricultural tradables in the model: coffee, soybeans and wheat. For processed agricultural products, as well as the non-agricultural products, on the other hand, there is imperfect substitution between domestic production and imports, which is modelled through a CES formulation which defines a composite good for those activities. The prices of urban domestic production are defined by costs of production, while import prices are dependent on exogenous external prices and tariffs. Urban export prices are 'made' internally, through domestic prices and taxes on exports. The demand for these goods is not perfectly elastic, though it can be high in some cases. In the agricultural sector, the production structure is specific to vegetable and animal products, both experiencing decreasing returns relative to an aggregate input, made up of urban and rural products, land and labour. The model seeks to keep track of complementarity and substitution patterns in the way primary factors and intermediates combine in the production process. This is done through a CBS-linked structure, specific to the types of production

3 Adjustment, Subsidies and Growth in Brazil 403 activity, with two levels of linkages and two substitution elasticities. This kind of structure divides the cost minimization problem into two sub-problems, separating inputs in each level of production from those in other levels. Production in the urban sector displays constant returns, through a Leontief formulation for intermediate inputs and a value-added aggregate that is a CES combination of capital and labour. The nominal urban labour wage is considered rigid, parameterized to reproduce the trend of urban unemployment in the period under analysis. Agricultural wages are flexible and adjust to clear the rural labour market in each period. The model has a neoclassical closure, in which total investment is given by saving. The nominal exchange rate is flexible and external capital flows are exogenous, so providing external sector closure. Imports, however, are subject to quantitative restrictions, in proportion to the desired imports. Government consumption is exogenous and public tax revenues endogenous, thus making the aggregate government current account endogenous. The closure of the government sector simulates an important mechanism used by the Brazilian government, in the relevant period, to finance its deficits, namely 'money creation'. The equilibrium between receipts and expenses is obtained through a type of 'seigniorage', that is, a variable that ensures the equilibrium in the government current account. This mechanism of transfer of funds between institutions is used to simulate the effects of a particular monetary phenomenon. CALIBRATION AND THE BASE RUN Having defined the theoretical model to be used the next step was its calibration. This was done through the construction of a social accounting matrix (SAM) for Brazil for There was a complication since the 1980 inputoutput matrix treated the agricultural sector as a single activity, although production was disaggregated. It was necessary, therefore, to split agriculture up into the 11 activities used in the study in a way described elsewhere (Ferreira Filho, 1996b). The resulting SAM for Brazil in 1980, evaluated at consumer prices, appears in Table 1 in a summarized form with products and activities aggregated to save space.2 This provides a broad outline of the nature of the magnitudes involved within the economy and need not be discussed in more detail. When the model was satisfactorily reproducing the observed pattern of the variables in the base year (1980) it was then run for the period 1981 to 1985, solving for the endogenous variables given the exogenous (observed) ones, so building the base run against which counterfactual analysis could be done. The model was solved as a non-linear optimization problem, using the General Algebraic Modelling System (GAMS) with MINOS5 (Brooke et al., 1988). The price of urban value added is the numeraire of the model.

4 404 Joaquim Bento de Souza Ferreira Filho TABLE 1 Social accounting matrix, Brazil, 1980 (millions of 1980 cruzeiros) Activities Products Labour agritrad agrinont agroind urbind agritrad agrinont agroind urbind Rural Activities agritrad agrinont agroind urbind Products agritrad agrinont agroind urbind Factors Rural labour Urban labour Rural capital Urban capital Institutions Rural families Urban families Government Capital account Stocks ROW Total expenses Notes: agritrad = agricultural tradables; agrinont = agricultural non-tradables; agroind =processed agricultural products; urbind =urban industry. EMPIRICAL RESULTS For the counterfactual analysis, the model was run again for the period, but the values of the agricultural subsidies were adjusted to fit the value observed in 1980, taken as 2.4 per cent of total GDP. The results were then compared with the base run. Table 2, for example, shows the results of the TABLE2 base run Macroeconomic aggregates as relative variations from the Year GDP EXP IMP CONS INV RERl RER

5 Adjustment, Subsidies and Growth in Brazil 405 Labour Capital Institutions Rural Urban Total Urban Rural Urban families families Government Capital Stocks ROW receipts I I I I experiment concerning some macroeconomic aggregates and the exchange rate. The values are shown in relation to the base run, being expressed in the form of relative variations (that is, multiplication by 100 provides percentage changes). Aggregates covered are GDP, exports (EXP), imports (IMP), consumption (CONS) and investment (INV), plus two concepts of real exchange rates. These are the nominal exchange rate (RERl), which can be seen as a real (deflated) exchange rate since the price of urban value added is the numeraire, while the other is deflated by an index price of domestic products (RER2). It can be seen that the volume of subsidies is large enough to cause macroeconomic effects in the model. It is interesting that the maintenance of agricultural subsidies at the levels observed in 1980 causes a general fall in macroeconomic aggregates in the model. To give one example, GDP could have fallen by as much as 4.4 per cent in 1984 and by 3.9 per cent in The 'increase' in agricultural subsidies, for that is what is implied, is also associated with a larger drop in aggregate investment, meaning that aggregate saving would have fallen, since, by the neoclassical closure of the model, investment is determined by the amount of saving. This happens because the income transfer to agriculture is made, in the main, through a reduction in urban income. The transfer is financed by

6 406 Joaquim Bento de Souza Ferreira Filho government through seigniorage, extracted from rural and urban sectors in proportion to the share of each in GDP. Thus the urban sector contributes the bulk of the agricultural subsidy. Since rural families have a greater marginal propensity to consume than urban families, the transfer affects consumption more than saving, thus reducing aggregate saving and investment. The government deficit (which is not shown in Table 2) would have considerably worsened in the period. Seigniorage needs would have risen from 4.23 to 4.95 and 9.5 per cent, respectively, in 1983, 1984 and 1985 (in the base run), and from 5.87 to 8.44 and 12.9 per cent (in the simulations), for the corresponding years. This, of course, should be interpreted as an indication that government income, as determined in the model, would not be enough to meet its expenses. The reduction in aggregate saving leads to a drop in investment, mainly because investment absorption is concentrated in some urban industries, notably manufacturing and services which have a high value-added coefficient in production. With the drop in investment, there is a fall in the demand for those composite products, reducing both domestic production and imports. As a consequence, the 'price' of domestic production falls and imports are reduced, resulting in the 'fall' in the equilibrium exchange rates (defined as Cr/US$, so that revaluation is involved). The changes in the composition of internal absorption, from high value-added activities to relatively low ones, is then associated with a fall in GDP (and then in imports). Turning now to variables more closely related to agriculture, the evolution of agricultural production and prices appear in Tables 3 and 4. It can be seen that only wheat, soybean and corn show important increases in production in the 'high subsidies' experiment. These are three of the four agricultural tradables in the model, along with coffee. As regards the latter, however, production would have fallen. That would have been the result because the agricultural production function is defined in terms of the relative prices of products and TABLE3 the base run Changes in agricultural production as relative variations from Year Coffee Sugarcane Rice Wheat Soybean Cotton Corn Other agric Poultry Livestock Milk

7 Adjustment, Subsidies and Growth in Brazil 407 TABLE4 base run Evolution of agricultural prices as relative variations from the Year Coffee O.Q Sugarcane Rice Wheat Soybeans Cotton Corn Other agric Poultry Livestock Milk inputs. For coffee, the fall in input prices, as a result of the subsidy, would not have compensated for the drop in its price caused by the revaluation in the exchange rate. So, despite the subsidy, the coffee/input relative price would have fallen, reducing production. For soybeans and wheat, on the contrary, relative prices would have been greater with heavier subsidization. Although their market prices would have been affected in the same proportion as that of coffee owing to the exchange rate effect, the rate of subsidy in the cost of the composite input in those cases is considerably higher than for coffee. The rate was about 7 per cent for coffee, but 32, 30 and 25 per cent for wheat in 1981, 1982 and 1983, respectively, and 18, 17 and 14 per cent for soybeans. Similar influences would have affected corn, although it is not a tradable product in the model, with production rising as a result of high subsidies on input prices. Cotton and sugarcane, which are also non-tradable products in the model, have their prices reduced. Since they are mainly inputs to the export agroindustries, their prices are strongly linked to exchange rate movements, though they benefit less from input subsidies. The three animal production activities, dealing with poultry, livestock and milk, would have experienced higher prices. This is a consequence of redistributive effects raising demand for those products, notably working through increasing rural disposable incomes, as shown in Table 5. In effect, the financing mechanism assumed to be adopted by government for the model (it mirrors what happened in the 1980s) is, in fact, a mechanism that redistributes income from the urban to the rural sector. The rise in the production of soybeans and wheat would have had a marked impact on external trade. Soybean exports would have risen substantially in value, while wheat imports would have fallen. However, raw coffee exports would also have risen, though not as a consequence of any increase in production. There would have been a decline in urban consumption, not matched by

8 408 Joaquim Bento de Souza Ferreira Filho TABLE 5 Changes in rural and urban disposable income as relative variations from the base run Disposable income Years Rural Urban expansion in rural consumption, allied to a fall in processed coffee exports caused by the revaluation of the equilibrium exchange rate. The maintenance of rural subsidies at levels observed in 1980 would have generated a strong rise in rural wages. As a result, their share in disposable income would have improved, with the positive variation reaching a maximum of 12 per cent, in This is shown in Table 6. In effect, rural wages would have appropriated a considerable share of the subsidies. It should be noted, however, that these results depend strongly on the hypothesis made about the evolution of the stock of agricultural machinery ('tractors'). Observed values for the 1980s showed a yearly rate of growth of 4 per cent, with that rate being used in the base run. But it has to be recognized that the size of the stock is not independent of the size of the subsidy programme. To explore the extent to which the results are affected, a rate of growth of 7 per cent, which was actually still below the growth rate of 10.3 per cent in the period , was inserted to obtain a new solution. The results appear in Table 7. TABLE6 run Value added, real wages and labour shares relative to the base Year Value added Real wages Rural labour share in agricultural rural urban rural urban disposable income The new results suggest that a higher rate of growth of mechanization would modify the earlier results. Although rural wages would still have risen in some years, the greater availability of tractors would have favoured substitution towards their use, generating a net fall in the rural wage share in total rural income.

9 Adjustment, Subsidies and Growth in Brazil 409 TABLE 7 Alternative scenario with tractor stock increasing at 7 per cent annually, variations from the base run Year Value added rural urban rural Real wages urban Rural labour share in agricultural disposable income CONCLUSIONS The CGE results provide interesting insights that would not have been obtained from partial equilibrium analysis. In a programme of agricultural subsidies, which is large enough to generate macroeconomic effects, the sources of funding are important. Though they are always dependent on the various hypotheses made about the structure of the economy, as well as on the parameter values chosen, CGE results make it possible to analyse the various links and feedbacks observed in a highly complex and interdependent economy. Perhaps of greater importance than the particular solution magnitudes for any of the variables is the demonstration of the possibility that a programme of agricultural subsidies can, in some circumstances, have unexpected results. Our results can help understand why the halting of the Brazilian agricultural subsidy programme in the 1980s, unexpectedly, did not have any significant impact on agricultural production. One popular explanation is that productivity gains would have offset the apparent fall in direct incentives in the relevant period. The results obtained with the Megabras model, however, suggest that other forces, working through market mechanisms in a general equilibrium situation, could also have played their part. NOTES 1The complete system of equations will be omitted, since it is too large. The interested reader should refer to Ferreira Filho (1995), for the complete system used, or consult Burniaux et al. (1990) for the basic structure of the RUNS model. Alternatively, the author can be contacted directly for additional information. 2The original disaggregated SAM can be obtained from the author upon request. REFERENCES Brooke, A., Kendrick, D. and Meeraus, A. (1988), GAMS: A user's guide, Washington, DC: The World Bank and The Scientific Press.

10 410 Joaquim Bento de Souza Ferreira Filho Bumiaux, J.M., van der Mensbrugghe, D. and Waelbroek, J. (1990), 'The food gap in the developing world: a general equilibrium modelling approach', in I. Goldin and 0. Knudsen (eds), Agricultural Trade Liberalisation: Implications for Developing Countries, Paris: OECD. Ferreira Filho, J.B.S. (1995), MEGABRAS- Um modelo de equilfbrio general computdvel aplicado ii andlise da agricultura brasileira, Sao Paulo: FENUSP, Doutorado. Ferreira Filho, J.B.S. (l 996a), 'Ajustamento estrutural e crescimento agrfcola na decada dos oitenta: notas adicionais', working paper, Department of Agricultural Economics and Sociology, Escola Superior de Agricultura Luiz de Queiroz, Universidade de Sao Paulo. Ferreira, Filho, J.B.S. (1996b), 'Um matriz de contabilidade social para o Brasil em 1980', working paper, Department of Agricultural Economics and Sociology, Escola Superior de Agricultura Luiz de Queiroz, Universidade de Sao Paulo. Shirota, R. (1988), Credito rural no Brasil: subsfdio, distribui{:tio efatores associados ii oferta, Piracicaba: Mestrado.